Different statistical methods are currently being used to analyze epidemiologic studies of diet and disease. This research project is examining the statistical properties of these methods. During the past year, a paper was written summarizing the advantages and disadvantages of different approaches to energy adjustment. Work has begun on new statistical models for incorporating measurement error in dietary intake, and the models are being tested on data from several dietary calibration studies. The Armitage-Doll mathematical model of carcinogenesis is being applied to experimental data on p53-knockout mice. Preliminary analyses indicate that inhibition of the normal p53 gene's tumor suppressor function by knocking out both alleles of the gene is consistent with 1 or 2 "stages" in the cancer process. Experimental data on hemizygotes with only one allele knocked out are being collected for additional analyses. New statistical methods for detection and inference of disease clusters are being developed. Previously proposed methods have been tests for overall clustering and do not have the ability to identify the location of clusters. The properties of a spatial scan statistic which takes into account the nonhomogeneous population densities as well as confounding variables have been evaluated. The statistic both tests for the location of clusters and tests for their statistical significance, and thus can be used both to evaluate cluster alarms and for routine surveillance. The spatial scan statistic has also been extended to study space-time clusters in addition to purely spatial ones.